Chris Dundon has a joke he tells on himself: “I don’t get to have fun anymore.” As an IT Applications Architect at Relativity, he’s moved from writing code to approving other people’s, evaluating every system that touches the company’s ecosystem, and asking the unglamorous questions nobody wants to answer twice. What kind of environments do you have? How do you handle permissions? What happens when something breaks?
For seven years, the answers to those questions came from a legacy cloud ETL platform. It worked, until it didn’t.
The slow leak
Relativity is a nearly 2,000-person legal data intelligence company, and, like most fast-growing organizations, its needs didn’t remain static. As the business added new systems, from CRM to billing to case management, its old integration platform started showing its age.
“We found the previous solution was deeply lacking in its ability to integrate outside of its comfort zone,” Chris said. “It could integrate with HTTP RESTful connections, but it was a heavy lift to do so.”
The cracks went deeper than the connectors. As Relativity expanded how it used its tools, Chris knew it needed to apply a coherent environment strategy to its middleware tooling. That was when the platform’s severe limitations became glaring: its environment strategy, its source control integration, and its documentation all led to dead ends.
“The platform had no environment strategy,” he said. “They claimed to have multiple environments, but they were impossible to use. I couldn’t figure out how to connect to source control, despite them claiming the ability to.” When he turned to the documentation for help, he found it scattered across a decade of different sites, most of it written for an on-premises version of the tool nobody at Relativity was using. “Getting information was impossible,” he said.
Meanwhile, the daily cost of that complexity kept compounding. Artem Milovanov, a Senior Middleware Engineer at Relativity, remembers one especially painful stretch: after a routine driver upgrade, every job connected to it had to be manually reviewed and reapproved, one at a time, across more than 100 jobs. “Approving a task that should have taken me 10 seconds instead required minutes, sometimes hours,” he said.
For a company racing to keep up with its own growth, that kind of friction doesn’t stay small. It piles up, quietly, until it’s a backlog no one can see the bottom of.
The breaking point
Every vendor relationship has a moment where the math stops working. For Relativity, it arrived at renewal time.
Getting it started with pricing that was hard to understand. The vendor’s pricing documentation was deeply nested and spread across multiple separate published documents and large reference spreadsheets. Amid all that complexity, it became easy to overlook that the vendor used a metric that was in no way an industry-standard measure of unique connections to a data warehouse. The result was a challenging renewal discussion.
That was the moment Relativity stopped casually browsing for alternatives and started actively building a replacement plan.
Choosing a platform, not just a tool
When Chris evaluates a major enterprise system, he doesn’t look at one option. He compares the incumbent against at least two serious alternatives, weighing connector pricing, environmental strategy, required skill level, and each vendor’s roadmap for the next 1 to 3 years. This time, that meant comparing staying put, looking into other vendors, and evaluating Workato.
Workato won, and not narrowly. Chris was able to fully author custom Ruby-based connectors himself, something that would have been a heavy lift on the old platform. “It’s been a very good experience,” he said.
Just as important, Workato didn’t require a developer to run it. Its visual, workflow-style interface meant non-technical Salesforce administrators could build and manage their own integrations, extending integration work well beyond Chris’s team, a deliberate goal, not an accident. “Workato has a workflow mentality,” Chris said. “You click through and design logic in a visually clean, simple interface.”
And the evaluation process itself became the safeguard against another ambush for renewal. Instead of discovering pricing surprises after signing, Relativity now runs every major tool decision through the same structured, transparent comparison, up front, before there’s a contract to be surprised by.
The math has borne that out. Since standardizing on Workato, Relativity has cut more than two weeks off the timeline for new integration projects, work that used to sit in a queue behind a handful of specialized developers now moves at the pace the business actually needs. Operational costs on its existing integrations have dropped by more than 30%, a world away from the six-figure overage the old vendor sprang on them at renewal.
A new kind of risk
Once the integration layer was solid, a different problem started to surface, one that had nothing to do with legacy debt and everything to do with what comes next.
As AI agents began appearing in enterprise software, vendors raced to ship MCP servers, giving AI models direct access to systems like Salesforce and Jira. Chris watched that rollout happen elsewhere and didn’t like what he saw.
“The thing with these enterprise MCP servers the vendors roll out is that the governance follows several months later,” he said. “That’s my biggest problem with vendor-based MCP servers. You don’t have the governance up front.”
He’s blunt about what’s at stake. Relativity’s Salesforce instance holds sensitive client and case data, plus years of automations nobody fully remembers building. Granting an AI model unrestricted create, read, update, and delete access without limits is not a hypothetical risk to him.
“It takes one misstep of the AI,” he said. “It’s trying to outthink you, trying to be smarter than your simple sentence, trying to anticipate your needs. I need some way to ensure we can prevent the system from going too far when it calls into a mission-critical system.”
So instead of accepting the vendor’s default, Relativity built its own MCP servers inside Workato, using role-based access and service account connections rather than broad, user-level permissions. The first one, a Jira Data Center MCP server, is live in production today, allowing users to search, create, comment on, and transition issues, all through the same secure, Okta-driven connection flow used across the platform. Gong and Gainsight are next.
Learning to trust the machine, carefully
Alongside MCP, Chris has been an early user of AIRO, Workato’s multi-agent platform for building recipes, and his skepticism has started to soften.
He needed to pull Salesforce permissions data into the warehouse for a routine audit report, and rather than think through the steps himself, he described what he needed to AIRO: pull two queries, send the results as two CSV attachments in a single email. “I had a working recipe a minute later,” he said. “It worked immediately.” By his own estimate, the same task would have taken him 30 to 60 minutes to build by hand.
What’s won him over isn’t just the speed. It’s what AIRO explains itself. It tells him what it did after it does it, and now asks permission before committing changes to a recipe at all, a small design choice that matters a great deal to someone whose entire job is making sure systems don’t do anything they weren’t supposed to. He’s started using it to learn unfamiliar connectors, too: build first, then compare what AIRO did with the documentation afterward, instead of reading the manual cover to cover before touching anything.
The throughline
It would be easy to read Relativity’s story as two separate chapters: a legacy migration and, years later, an AI story. But they’re the same instinct, applied twice. Chris evaluates every system the way an architect evaluates infrastructure: does it fit the ecosystem, does it scale with the skill level of the people who’ll actually use it, and does it come with guardrails before something goes wrong, not after.
The old ETL platform failed that test on cost and capability. Vendor MCP servers were failing on governance. Workato is the one system that’s passed it twice so far.
